Adaptive neuro-fuzzy inference system (ANFIS) for prediction the gibbs energy of formation

The Gibbs free energy of formation is very important property to predict the feasibility of chemical reactions, calculate equilibrium constants, and analyze reaction spontaneity. In this work, a nonlinear group contribution-based method was developed by combination of adaptive neuro-fuzzy inference system (ANFIS) and group contributions (GC) methods. In order to obtain simple and accurate model for prediction the gibbs free energy of formation of acid compounds, the genetic function approximation (GFA) method was used for selection the most important functional groups. The required parameters of the models are the numbers of occurrences of three functional groups in each investigated acid molecule, which can be computed based on chemical structure of any acid molecule. Using the ANFIS method the gibbs free energy of formation predicted with high accuracy and reliability that are quantified by the following statistical parameters: the squared correlation coefficient (R²) = ۰.۹۷۸, absolute average relative deviation percent (AARD %) =۳.۹۷۷, and the root mean squares error (RMSE) = ۲۶.۹۳۲. In general, the results obtained in this work showed that the ANFIS-GC could be a promising machine learning approach to predict the gibbs free energy of formation, or possibly other physiochemical properties of acid compounds.

کلیدواژه ها:

Acid ، Adaptive neuro-fuzzy inference system (ANFIS) ، Genetic function approximation (GFA) ، Gibbs free energy of formation ، Group contribution (GC) ، Machine learning